Evaluating Somfit’s pulse arterial tonometry for detection of obstructive sleep apnoea

Abstract

AbstractThis study evaluates the diagnostic accuracy of Somfit against polysomnography (PSG) for detecting obstructive sleep apnoea (OSA). Somfit is a wearable home-sleep monitoring device attached to the forehead, combining pulse arterial tonometry, oximetry, and actigraphy with sleep staging, arousals, and total sleep time (TST) derived from frontal neurological signals. Ninety-two participants suspected of having OSA were assessed using Somfit and simultaneous overnight PSG recordings at three Australian sites. Each PSG study was manually scored by three independent scorers. The reported statistics include standard measures of agreement between Somfit’s TST, Oxygen-Saturation Index (ODI), Apnoea–Hypopnea Index (AHI), and the average of those metrics from the three PSG scorers. The overall inter-scorer agreement was 76% (kappa = 0.772). TST, ODI, and AHI from Somfit were highly correlated with similar metrics from PSG (all r > 0.84, p < 0.001). Sensitivity, specificity, and accuracy were 90.5%, 75.0%, and 89.1% respectively, with a diagnostic odds ratio (DOR) of 28.5 for AHI ≥ 5. For AHI ≥ 15, sensitivity, specificity, and accuracy were 78.0%, 73.8%, and 76.1%, respectively, with a DOR of 9.99. For AHI ≥ 30, sensitivity, specificity, accuracy, and DOR were 72.4%, 90.5%, 84.8%, and 24.9, respectively. The area under the curve (AUC) at different PSG AHI cut-offs ranged between 0.86 and 0.93. Coupled with oximetry and EEG-based derivation of TST, Somfit’s performance is comparable to PSG in detecting OSA severity.